Sequencing of multi‑robot behaviors using reinforcement learning

Given a collection of parameterized multi-robot controllers associated with individual behaviors designed for particular tasks, this paper considers the problem of how to sequence and instantiate the behaviors for the purpose of completing a more complex, overarching mission. In addition, uncertainties about the environment or even the mission specifications may require the robots to learn, in a cooperative manner, how best to sequence the behaviors. In this paper, we approach this problem by using reinforcement learning to approximate the solution to the computationally intractable sequencing problem, combined with an online gradient descent approach to selecting the individual behavior parameters, while the transitions among behaviors are triggered automatically when the behaviors have reached a desired performance level relative to a task performance cost. To illustrate the effectiveness of the proposed method, it is implemented on a team of differential-drive robots for solving two different missions, namely, convoy protection and object manipulation.     

Effect of artificial aging on intergranular corrosion of extruded AlMgSi alloy with small Cu content

The effect of artificial aging parameters on the corrosion performance of air cooled AlMgSi(Cu) model alloy extrusions was investigated. Accelerated corrosion test revealed that the extrusions were highly susceptible to intergranular corrosion (IGC) in the naturally aged condition. However, IGC susceptibility was reduced, and finally eliminated, by artificial aging. Overaging introduced slight pitting susceptibility. EDS X-ray mapping in FE-TEM revealed Mg₂Si and Q-phase (Al₄Cu₂Mg₈Si₇) grain boundary precipitates and a continuous Cu-enriched grain boundary film. IGC susceptibility was related to the Cu-enriched grain boundary film. Increased IGC resistance was caused by coarsening of the grain boundary film by aging. Pitting susceptibility by over aging evolved due to coarsening of the Q-phase particles in the grain bodies.     

Experimental and numerical life prediction of thermally cycled thermal barrier coatings

This article addresses the predominant degradation modes and life prediction of a plasma-sprayed thermal barrier coating (TBC). The studied TBC system consists of an air-plasma-sprayed bond coat and an air-plasma-sprayed, yttria partially stabilized zirconia top layer on a conventional Hastelloy X substrate. Thermal shock tests of as-sprayed TBC and pre-oxidized TBC specimens were conducted under different burner flame conditions at Volvo Aero Corporation (Trollhättan, Sweden). Finite element models were used to simulate the thermal shock tests. Transient temperature distributions and thermal mismatch stresses in different layers of the coatings during thermal cycling were calculated. The roughness of the interface between the ceramic top coat and the bond coat was modeled through an ideally sinusoidal wavy surface. Bond coat oxidation was simulated through adding an aluminum oxide layer between the ceramic top coat and the bond coat. The calculated stresses indicated that interfacial delamination cracks, initiated in the ceramic top coat at the peak of the asperity of the interface, together with surface cracking, are the main reasons for coating failure. A phenomenological life prediction model for the coating was proposed. This model is accurate within a factor of 3.     

Assessing Decentralised Water Solutions: Towards a Framework for Adaptive Learning

This paper reports on the use of qualitative analysis to inform a risk analysis framework for decentralised water systems. To realise the benefits from these technologies, a methodology is applied to learn from previous difficulties in implementing and managing them. A workshop process was used to capture stories from industry professionals on difficulties they have encountered in planning and implementation. Qualitative analysis of story narratives revealed stages where there was some type of development process failure; as well as failure modes and factors influencing the difficulties encountered. The analysis also generated insights: difficulties in one part of the development process tends to propagate to subsequent stages; system difficulties most often occurred in the policy stage of development due to institutional inertia and lack of adaptive governance; and the best indicator of problems with a decentralised system was complaints of poor water quality. Furthermore, this paper also provides a method to learn from past difficulties by identifying what data needs to be collected in order to populate a risk model which can be used for improving risk assessment of the development process for decentralised systems. This can provide a basis for better decision making, policy and guidelines; an important factor in mainstream acceptance.     

Application of particle swarm optimization in epileptic spike EEG source localization

Surgical therapy has become an important therapeutic alternative for patients with medically intractable epilepsy. Correct and anatomically precise localization of an epileptic focus is essential to decide if resection of brain tissue is possible. The inverse problem in EEG-based source localization is to determine the location of the brain sources that are responsible for the measured potentials at the scalp electrodes. We propose a new global optimization method based on particle swarm optimization (PSO) to solve the epileptic spike EEG source localization inverse problem. In a forward problem a modified subtraction method is proposed to reduce the computational time. The good accuracy and fast convergence are demonstrated for 2D and 3D cases with realistic head models. The results from the new method are promising for use in the pre-surgical clinic in the future.     

Agent based simulation architecture for evaluating operational policies in transshipping containers

An agent based simulator for evaluating operational policies in the transshipment of containers in a container terminal is described. The simulation tool, called SimPort, is a decentralized approach to simulating managers and entities in a container terminal. Real data from two container terminals are used as input for evaluating eight transshipment policies. The policies concern the sequencing of ships, berth allocation, and stacking rule. They are evaluated with respect to a number of aspects, such as, turn-around time for ships and traveled distance of straddle carriers. The simulation results indicate that a good choice in yard stacking and berthing position policies can lead to faster ship turn-around times. For instance, in the terminal studied the Overall-Time-Shortening policy offers fast turn-around times when combined with a Shortest-Job-First sequencing of arriving ships.     

Digital smoothing of the Langmuir probe I-V characteristic

Electrostatic probes or Langmuir probes are the most common diagnostic tools in plasma discharges. The second derivative of the Langmuir probe I-V characteristic is proportional to the electron energy distribution function. Determining the second derivative accurately requires some method of noise suppression. We compare the Savitzky-Golay filter, the Gaussian filter, and polynomial fitting to the Blackman filter for digitally smoothing simulated and measured I-V characteristics. We find that the Blackman filter achieves the most smoothing with minimal distortion for noisy data.     

Detailed 3D human body reconstruction from multi-view images combining voxel super-resolution and learned implicit representation

Applied Intelligence - The task of reconstructing detailed 3D human body models from images is interesting but challenging in computer vision due to the high freedom of human bodies. This work...